Bug#24717: APT robustness
I hate it when "advocacy" comments appear on bugs but I'm going to do this
anyways. This isn't a minor bug, I wouldn't even consider this a wishlist bug,
this is a fundamental and essential piece of logic before APT can be
considered to be working properly.
> On Fri, 17 Sep 1999, Adrian Bridgett wrote:
>
> > I'd like to be able to continue - if _one_ package is broken, then it can
> > hold up installing all the rest. Is there a particular reason for the stop?
>
> There are a few bug reports with discussions on this topic. The end result
> is that dpkg does not provide adaquate information to even deterimine
> which package failed so there is no way to work around the problem.
>
> Jason
This is BS. You don't need any such feedback from Dpkg. All the information
you need is in the package's current status. The broken package won't be
configured and therefore will cause dependency problems as necessary with
subsequent packages.
The brute force way to implement this is to stop, put any installed but not
configured packages on hold, then recalculate everything that APT would have
done for the original operation all over again.
This is basically equivalent to the same operation the user is performing by
repeatedly running APT and manually holding packages that are failing to
configure.
--
greg
Reply to: